Structural Similarities Between Language Models and Neural Response Measurements
This work addresses the problem of understanding how artificial language models relate to human brain processing, which is incremental in linking AI and neuroscience.
The study investigated the relationship between the internal representations of large language models (LLMs) and neural response measurements from brain imaging, finding that larger LLMs exhibit greater structural similarity to these neural representations.
Large language models (LLMs) have complicated internal dynamics, but induce representations of words and phrases whose geometry we can study. Human language processing is also opaque, but neural response measurements can provide (noisy) recordings of activation during listening or reading, from which we can extract similar representations of words and phrases. Here we study the extent to which the geometries induced by these representations, share similarities in the context of brain decoding. We find that the larger neural language models get, the more their representations are structurally similar to neural response measurements from brain imaging. Code is available at \url{https://github.com/coastalcph/brainlm}.